Methodology

Secret Stocks exists to bring institutional-grade fundamental analysis to retail investors. We apply the same methodology to every company we cover, expose how the analysis is generated rather than hiding it behind a black-box score, and aim to make decades of financial history legible at the scale of the entire S&P 500. We are a publisher of automated, impersonal analysis — not an adviser.

Financial Modeling Prep (FMP)is our single source of financial data. We pull raw statements (income statement, balance sheet, cash flow), historical market capitalization, end-of-day prices, and company profiles from FMP’s /stable/ endpoints, then compute valuation multiples ourselves on a Trailing-Twelve-Months basis to preserve historical accuracy.

SEC EDGAR is the gold standard for verification. Where a number matters to your decision, we strongly recommend cross-checking it against the original filings at EDGAR full-text search. Our platform does not replace primary filings; it summarizes third-party data that may itself contain errors.

Content is produced by an orchestrated, automated pipeline. We use n8n Cloud as the orchestration layer, Google Gemini Flash for standard metric summaries and Gemini Pro for deeper analysis, and OpenAI gpt-image-1 for featured images.

Every article passes through a human approval gate before publication. The reviewer performs a plausibility check — verifying that outputs are not malformed, offensive, or severely hallucinated — and rejects clearly defective content. The reviewer is not a licensed financial professional, does not verify every data point against original filings, and does not perform editorial quality assurance comparable to traditional financial publishing. This gate is a safety net against catastrophic AI failure, not a guarantee of accuracy.

• AI hallucinations: LLMs can fabricate figures, conflate companies, or invent events that never occurred.
• API latency & delayed data: Third-party data may be stale, revised, or temporarily unavailable.
• LLM arithmetic errors: Language models are unreliable calculators; derived numbers should be treated as estimates.
• Concept drift: Model updates can change tone, behavior, and quality over time without notice.
• Shallow approval depth: The approval gate catches catastrophic failures, not subtle inaccuracies.

Treat everything here as a starting point, not a conclusion. Before acting:

• Cross-check SEC EDGAR:Confirm headline figures against the company’s actual 10-K / 10-Q filings.
• Confirm across multiple sources: Do not rely on a single number from a single provider.
• Be skeptical of extraordinary numbers: If a metric looks too good (or too bad) to be true, assume a data or model error until proven otherwise.

• SEC Publisher’s Exclusion: We operate as an impersonal publisher under Section 202(a)(11)(D) of the Investment Advisers Act of 1940, not as a registered investment adviser.
• EU MAR Article 3(1)(35): Where outputs qualify as investment recommendations, we present them objectively and disclose conflicts of interest.
• EU AI Act Article 50: AI-generated content is clearly disclosed as such at the point of publication.

For the full risk and liability language, see our Disclaimer and Terms of Service.